Alternate Exits?
you could easily make a maze this way by maybe having the obvious finish of a series of levels take you back to a level you already did, and have the 'secret' or alternate exit in one of the levels be the real exit, do this many times, so it would be challenging to find the real exit. this way it would be unlikely for people to miss content, and forcing the player to replay levels or not find the exit might help reinforce any storyline you have. portal did it. and mazes are fun!
Some of your argument seems true, but there are some flaws to your logic.
First, by a map maker's standpoint, If you try to put as many different maps as there were in portal into a SINGLE map file, Hammer would hate you. Keeping things organized with 20 different levels, each with many rooms would be quite the challenge. Trying to keep track of hundreds of entities without using the same entity name twice... not a fun time. Also, if you didn't keep the rooms in their own vis groups... Hammer would explode trying to show you every brush you made. Another problem with having one giant map is it could (potentially) limit the design of a map, if you want a room to go back into a space already occupied by another room.
Also, compiling would take ages, as you wouldn't be able to only re-compile the one room you are changing. If you read some articles about map optimization it will go in to visleafs and portals... but, it would just take forever to compile the whole new map, after you moved one slanted ramp up 32 units to see if the puzzle will work now.
Second: The end-user's side. Simply put, that one single load time would suck.
/endrant.
for the first few levels of portal there are two chambers to one file, for the later, larger chambers they are one each, they are quite large compared with how they look in-game
and lots of little loads beats one huge jumbo load.
Compile time?
lol sorry you guys have it easy today with how fast maps compile. When I worked on maps for dod 1.3b it would take a solid 2 hours to fully compile a map
Also I dont know about you guys but I really do not like to have to go backwards in maps and do things again. I want to go to new areas and see new things not get bogged down doing 52 things in a small space. I would say no more than 4 things for a large room and done all at the same timeframe.
Compile times are relative to the complexity of the map. Portal maps are, almost by definition, fairly simplistic.
Please do not Private Message me for assistance. Post a thread if you have questions or concerns.
If you need to contact the staff privately, contact the Global Moderators via Discord.
I really want to do a level set where you go through two or three test chambers normally, then find a way "behind the walls" and go back through the office areas and "backstage" areas (cube delivery etc) of the chambers you just did.
Valve did this a little bit in the single-player, but not to the extent I'd like to do it.
To the folks arguing about how it's a bad idea:
It seems very few of you have played the original DOOM or remember how it worked if you have.
first of all, missing content?
The whole point of the secret levels was to SEE the content.
Basically you had this path assuming start room "a", end room "b" and
secret room "c".
a -> b
or
a -> c -> b.
it worked by making the "normal" exit in "a" take you to "b", and
the "secret" exit took you to "c" which in turn typically only had one exit
for purposes of going back to "b" afterwards. You didn't necessarily have
to make an entire "conditional" tree where one exit makes it branch
COMPLETELY away from the rest of the maps. They actually would lead
BACK to the main map. Which is the whole point of simulating non-linearity aka
"multiple paths" without FULLY implementing it.
Very few games even go that route anyway as the whole "conditional"
thing is difficult to complete anyway.
secondly, all the maps were always separate files, and it would only
make sense to do so. The maze idea seems great, but if you can't
find that "special" exit eventually, it's going to get redundant and
frustrating from a gameplay experience. The whole idea of the "secret"
level was for bonus content that "could" be missed, but if you wanted to
be REALLY in depth with it, you could. (ie if you were just speedrunning it,
you can skip it obviously, if you wanted 100%, you may want to complete it...
That kind of idea).
Anyway, just my opinions on this, and if I ever get around to completing
at least ONE level, I'll eventually make a mappack with that style of
gameplay in mind....
This idea sounds cool
I would only have to ask, why spend so much time developing a level that not all the players will see because its "secret"?
I would think it would make more sense to just keep it linear, and have all your puzzles played.
Now one thing I can see, is having "secret" areas of a map, to change the intended solution.
or a button that will turn the whole room upside down
it can't be too secret or its not worth doing, but it should require thinking with portals to uncover 
secret room "c".
a -> b
or
a -> c -> b.
it worked by making the "normal" exit in "a" take you to "b", and
the "secret" exit took you to "c" which in turn typically only had one exit
for purposes of going back to "b" afterwards.
i'm still pushing the maze idea strong, somepeople might not like labyrinths, and that's ok. but most of the time, i appreciate a map more when played for the second time and it seems portal creators understand this too, since they make you play some levels more than once, then over and over in the "bonus maps" that aren't new at all.
also the longer you keep a player in your maps, the more value you get out of the hours you put in. portal maps it seems are one time use...
but by making a player go
a->b->a->c->end
maybe they would have to look for clues in 'a' that point them to 'c' that they missed the first time trying to get to the obvious exit.
i would really enjoy playing a map set like this.
I like the idea as it pertains to difficulty. Access to the secret areas could only be attained by solving some very difficult puzzles. The presence of an alternate exit doesn't even need to be hidden, so there's little worry that players will miss content without realizing it. The bonus content could then be significantly more difficult than the standard content, satisfying the hardcore portal braniacs without penalizing the more casual player who may not want that sort of gameplay anyway.